Open Standards for AI Content Transparency
Two standards for AI agent identity and content attribution. Compatible with RSL and other licensing standards. Apache 2.0 licensed.
Two standards for AI agent identity and content attribution. Compatible with RSL and other licensing standards. Apache 2.0 licensed.
AI systems increasingly rely on content created by others, yet there is no standard way to track, verify, or compensate for that usage.
When an AI agent requests content, publishers have no standard way to verify who it is, what it's licensed to access, or whether it will respect licensing terms. The choice is binary: block all AI agents or allow all of them.
AI systems consume articles, reviews, and product data, then present synthesized responses without attribution. Content creators get no visibility into how their work is used or what value it creates.
Even agents that want to operate transparently have no standard way to prove compliance. Publishers who license content have no tools to verify that agreements are being followed.
Publishers who block AI bots find their content scraped by third parties, repackaged on aggregator sites, and ingested by AI systems anyway. Without provenance standards, legitimate and laundered content are indistinguishable.
An adversarial relationship between AI systems and content creators. Publishers block AI entirely or get scraped without accountability. AI operators cannot prove compliance even when they want to. Users have no idea what's behind the AI they're talking to. The value created by content flows to those who aggregate it, not those who create it.
OpenAttribution defines two complementary standards for AI content transparency.
Identity & Compliance
AI Manifest Standard. DID-based agent identity with a three-layer manifest: foundation (training provenance), deployment (operator context), and content access (licensed sources). Includes disclosure headers for runtime compliance verification.
Behavior
Content Attribution Telemetry. Session-based tracking of content events (retrieved, cited, displayed), conversation turns, and commerce outcomes. Configurable privacy levels.
OpenAttribution is compatible with source-side licensing standards like RSL. Publishers define permissions on their content; OpenAttribution provides the consumer-side identity and tracking so agents can prove compliance.
| Standard | Question | Side |
|---|---|---|
| Source-side licensing (e.g. RSL) | What can AI do with this content? | Source (publisher) |
| AIMS | Who is this AI and what can it access? | Consumer (agent) |
| Telemetry | What did this AI actually do? | Behavior (audit) |
See which AI bots can access any website. Check robots.txt policies, RSL licenses, and identify blocked crawlers instantly.
Track GPTBot, ClaudeBot, Gemini, and more
Check single URLs or bulk CSV uploads
Download results as JSON or CSV
The full architecture from content source to telemetry destination.
Publishers use RSL to declare machine-readable licensing terms: what AI systems can do with their content, what compensation is required.
The agent publishes an AIMS manifest (DID-based identity + licensed sources) with a disclosure header on content requests, and emits Telemetry on content usage and outcomes.
Publisher resolves the AIMS manifest via the DID, checks that the agent is licensed for the requested content, and grants or denies access based on verification.
Content events (retrieved, cited, displayed), conversation turns, and outcomes are tracked. Publishers get usage reports. Compliance is auditable. Compensation can flow based on measured value.
OpenAttribution is built for multiple stakeholders across the AI content ecosystem.
pip install openattribution-telemetrypip install openattribution-aimsPython SDKs for both AIMS and Telemetry. Apache 2.0 licensed.
pip install openattribution-aimsfrom openattribution.aims import AIManifest
from openattribution.aims.layers import (
DeploymentLayer,
ContentAccessLayer,
)
manifest = AIManifest(
did="did:aims:web:example.com:my-agent",
deployment=DeploymentLayer(
operator="Example Corp",
specialization="customer_service",
),
content_access=ContentAccessLayer(
licensed_sources=["source:internal-docs"],
rsl_compliance=True,
),
)pip install openattribution-telemetryfrom openattribution.telemetry import Client
async with Client(
endpoint="https://telemetry.example.com",
api_key="your-api-key",
) as client:
session_id = await client.start_session(
content_scope="shopping-assistant",
manifest_ref="did:aims:web:example.com:agent",
)
await client.record_event(
session_id=session_id,
event_type="content_retrieved",
content_id=content_id,
)
await client.end_session(
session_id=session_id,
outcome=SessionOutcome(type="conversion"),
)Honest maturity levels for each component.
| Component | Status | Notes |
|---|---|---|
| AIMS Specification | Draft v0.1 | Content transparency layers complete; crypto sections incomplete |
| AIMS Python SDK | Alpha | pip install openattribution-aims |
| Telemetry Specification | Preview v0.3 | Comprehensive but not finalized |
| Telemetry Python SDK | Alpha | pip install openattribution-telemetry |
| Telemetry Reference Server | Alpha | Self-hostable reference implementation |
| AIMS Registry | Roadmap | Design partnership needed |
| Framework Integrations | Roadmap | LangChain, LlamaIndex, MCP priority |
| OpenTelemetry Export | Roadmap | OTEL semantic conventions proposal |
Open standards governed by a multi-stakeholder community.
All specifications and SDKs are Apache 2.0 licensed. Free to use, modify, and distribute.
Company Limited by Guarantee. No shareholders. Governed by members for the benefit of the standards.
Input from publishers, AI operators, platforms, and civil society. No single company controls the standards.
Publishers
Media, news, editorial
Brands
Retailers, manufacturers
Tech Providers
Attribution platforms, CDNs
AI Providers
LLMs, agent platforms
Agencies
Marketing, media services
OpenAttribution is an open standards body developing two specifications for AI content transparency: AIMS (agent identity and compliance) and Telemetry (content usage tracking). All specifications are Apache 2.0 licensed and community-governed.
OpenAttribution is compatible with source-side licensing standards like RSL (Really Simple Licensing). Publishers use standards like RSL to declare what AI can do with their content. OpenAttribution provides the consumer side — agents use it to declare who they are, what they're licensed to access, and track what they actually do. Together they create a complete licensing and compliance chain.
No. OpenAttribution provides infrastructure for AI systems to access content transparently and sustainably. AI systems need quality content. Content creators need measurement and attribution. These standards make both possible without requiring either party to trust the other blindly.
Not necessarily. While the SDKs are developer-focused, OpenAttribution needs input from publishers, brands, legal experts, and policy makers to shape the standards. Governance participation, pilot programs, and specification feedback are all valuable contributions.
OpenAttribution does not prevent scraping, provide cryptographic provenance for text, solve the "who said it first" problem, or guarantee that declared access rights are accurate. It provides a framework where those who want to operate transparently can do so, and recipients can distinguish between verified and unverified sources.
Install the Python SDKs: pip install openattribution-aims for agent identity and pip install openattribution-telemetry for content tracking. See the Developer Quick Start section for code examples. All code is on GitHub.
Interested in contributing to the standards? Get in touch.